Indirect Prompt Injection in AI Coding Assistants via Third-Party Source Code Leads to Credential Exfiltration | AI Breaking Wire